Table 2

Results of the hierarchical multiple regression analysis

 Predictor variables Estimate SE β* t Value p Value Model 1 Constant 0.31 0.003 – 114.28 <0.0001 Log proportion of Hispanic 0.91 0.01 0.29 68.52 <0.0001 Log proportion of black 0.43 0.01 0.17 41.74 <0.0001 Log proportion of families living in poverty 0.83 0.02 0.16 34.29 <0.0001 Log proportion of women older than 25 years without a high school diploma 0.34 0.02 0.09 17.33 <0.0001 Log average household size −0.57 0.01 −0.18 −54.02 <0.0001 Urban (1=urban, 0=rural) 0.32 0.003 0.33 93.90 <0.0001 Model 2 Constant 0.26 0.003 – 81.38 <0.0001 Log proportion of Hispanic 0.11 0.03 0.03 4.12 <0.0001 Log proportion of black 0.10 0.02 0.04 4.13 <0.0001 Log proportion of families living in poverty 0.75 0.05 0.15 15.11 <0.0001 Log proportion of women older than 25 years without a high school diploma 0.05 0.04 0.01 1.33 0.183 Log average household size −0.26 0.02 −0.08 −12.53 <0.0001 Urban (1=urban, 0=rural) 0.36 0.004 0.37 97.97 <0.0001 Urban by log proportion of Hispanic 1.01 0.03 0.29 33.16 <0.0001 Urban by log proportion of black 0.43 0.03 0.15 16.65 <0.0001 Urban by log proportion of families living in poverty 0.03 0.06 0.005 0.50 0.617 Urban by log proportion of women older than 25 years without a high school diploma 0.33 0.04 0.08 7.73 <0.0001 Urban by log average household size −0.44 0.02 −0.13 −18.35 <0.0001
• R2=0.39 for Step 1, ∆R2=0.02 (p<0.0001).

• The column header Estimate is the unstandardised parameter estimate. For all predictor variables except the binary urban/non-urban variable, the estimate represents the per cent change in the tobacco outlet density per 1000 population for a 1% change in the predictor variable since both the predictor and the outcome variables are log transformed. In the case of the binary urban/non-urban predictor variable, which is not log transformed, the estimate is the (100 × estimate) per cent change in the outcome variable for a unit change in the predictor variable. For the interaction terms, the interpretation is the additional percentage increase for urban census tracts versus rural census tracts. Thus, for instance, a 1% increase in the proportion of Hispanics in non-urban census tracts is associated with a 0.11% increase in the tobacco outlet density per 1000 population. For urban communities, there is an additional 1.01% change totalling to a 1.12% increase for a 1% increase in the proportion of Hispanics compared with a 0.11% increase in non-urban census tracts.

• * Standardized parameter estimate.